AI agents
Define an agent once. Your team gets it everywhere.
Platform specialists ship in every organisation. Add your own with explicit tool permissions. The same agent answers in chat, in Slack, in the terminal, and through the API.
plan finished its design. Handing off to code.
Included
Specialists in every organisation
You do not start from an empty list. Six specialist agents exist from day one, alongside an orchestrator and a GitHub operations agent. On QuantGov, the specialists carry the built-in ISM and Essential Eight skills.
Writes and edits code. In the terminal, it runs your tests too.
Investigates and proposes a design. Reads only, never edits.
Threat modelling and compliance assessment.
Code review with quality and compliance checks together.
Traces root causes before proposing a fix.
Researches dependency CVEs and flags unpinned or compromised packages.
Overlays
Customise without forking
A platform agent almost fits, but your organisation has its own conventions. Fork it and you inherit the maintenance. Leave it alone and it never quite matches how you work.
An overlay is the third option. It layers your changes on top of the platform agent — add to the prompt, change the model, adjust the tools, disable specific skills. Improvements we make underneath still reach you.
- Add your conventions to the system prompt
- Narrow or extend the tool allowlist
Disable skills you do not want that agent reaching for
- Keep receiving upstream improvements
plan finished its design. Handing off to code.
Tools
Tools your agents can actually call
A built-in tool library ships with every organisation. It includes 19 GitHub operations — read files, list commits, review and merge pull requests, manage issues and labels — so an agent works a repository the way a colleague does.
Define your own as edge functions and deploy them from the dashboard.
Register remote MCP servers and attach them per agent. Credentials stay in the gateway and are injected server-side, so they never reach a laptop or a browser. Tool schemas are hashed and pinned, so a server that changes its tools is detected and logged.
MCP documentation
Credentials injected here. Never sent to clients.
Surfaces
The same agent, wherever the work is
An agent you define is available to your whole team the moment you save it. It answers in the dashboard, responds to a Slack mention, runs in a developer’s terminal, reviews a pull request from a webhook, and executes a scheduled workflow overnight.
You do not rebuild it per surface, and the audit trail follows it between them.
What teams build
Agents are most useful when everyone reaches for the same one.
Repository review
A review agent with the GitHub tools and your standards as a skill. Wire it to a webhook and every pull request gets the same treatment.
- 19 GitHub tools
- Your standards as a skill
- Webhook triggered
Service desk
A narrow agent with knowledge base access and a short tool list, exposed in Slack where people already ask questions.
- Knowledge base access
- Narrow tool list
- Slack native
Compliance checks
The security specialist with an overlay for your obligations. Ask it directly, or put it in a workflow before release.
- Built-in compliance skills
- Overlay for your rules
- Runs in workflows
Give your team agents worth sharing
Start with the platform specialists and add your own.